PHP + MySQL'数据库连接'减慢服务器速度

Question

I have an ajax-loaded script which runs 2 mysql statements for every single visitor (~100 per second). It's a simple SELECT as well as UPDATE statement.

For my backend, I'm using AWS RDS.

I noticed my site went down today, I see the logs filled with the following...

PHP-FPM-5.5 Error log:

[11-Jun-2014 17:18:47] WARNING: [pool www] child 12203, script '/var/www/track.php' (request: "GET /track.php") executing too slow (5.282239 sec), logging

Slow log

script_filename = /var/www/track.php. [0x00007fcf11fda8f8] execute() /var/www/html/track.php:53

In RDS, my graphs looked like this:

You can see the DB connections became very high.

At this point is where I'm a bit confused what to do to prevent this.

I can only assume something 'backed up', maybe the database server took just a fraction of a second too long to respond? Then PHP basically got overloaded because the requests were coming in too fast (too busy 'waiting'), which in turn started to take down my server.

The MySQL statements I have are using PHP-PDO . I attempted adding PDO::ATTR_TIMEOUT => 1 to my new PDO object, but I have no idea if that would even solve it (or even if it's working or not...).

My goal is to NOT have my server get taken down just because the DB server is slow or unresponsive. I need PHP to be 'smart enough' to not wait for a really long time or get overloaded because MySQL isn't responding fast enough.

All this is pure conjecture though, maybe it wasnt RDS at all, and somehow PHP got backed up? I'm not exactly sure...

Thanks!

Answer 1

I'm using MySQL in my most recent project but can't deny the fact that MySQL may have some performance issues.

You could try to send a select request to your database and if it returns true you can then update the row. In this case you'll need two connections per user and per page load. So I will recommend that you use stored procedures which will only require one connection from your script. You could half your connection times with using stored procedures in your scenario.

My second solution is to configure your my.conf according to your required settings. Here is an online tool to help make a suitable configuration https://tools.percona.com/wizard .

MySQL can use up to 40% of system resources for logging so try to limit some logs such as access log, error log, and any additional logs you may be using.